nvmecontrol: Use admin device for admin-passthru
Needs ReviewPublic
Actions

Authored by imp on Mon, Sep 21, 4:57 PM.

Details

Reviewers

seuros
mav
chuck
chs
ken

Group Reviewers

manpages

Summary

When we're doing admin-passhtru, we need to use the admin device, not
the namespace device. So when we're passed in ndaX or nvdX, we need to
translate that to nvmeY for this command for the admin-passthru command,
but not the io-passthru command. Note namespace nuance in man page.

Diff Detail

Repository

rG FreeBSD src repository

Lint

Lint Skipped

Unit

Tests Skipped

Build Status

Buildable 77141
Build 74024: arc lint + arc unit

Event Timeline

imp created this revision.Mon, Sep 21, 4:57 PM

Herald added subscribers: ziaee, dab. · View Herald TranscriptMon, Sep 21, 4:57 PM

imp requested review of this revision.Mon, Sep 21, 4:57 PM

Harbormaster completed remote builds in B77141: Diff 187348.Mon, Sep 21, 4:57 PM

imp added a parent revision: D59835: nvmecontrol: add format dry-run support.Mon, Sep 21, 4:58 PM

imp added reviewers: seuros, mav, chuck, chs, ken.

ziaee added a reviewer: manpages.Mon, Sep 21, 5:10 PM

ziaee added inline comments.

sbin/nvmecontrol/nvmecontrol.8
769	This line was 85 characters and will whine the linter.

rpokala added a subscriber: rpokala.Mon, Sep 21, 5:25 PM

rpokala added inline comments.

sbin/nvmecontrol/nvmecontrol.8
769	Will the `.Pq` capture the terminal `.`, or will that remain outside the parens?

ziaee added inline comments.Mon, Sep 21, 6:46 PM

sbin/nvmecontrol/nvmecontrol.8
769	It will go outside. Also, it should never be inside. Parens is like a subshell inside a sentence.

chuck added inline comments.Mon, Sep 21, 11:18 PM

sbin/nvmecontrol/nvmecontrol.8
769	Looking at Figure 141: Opcodes for Admin Commands in the 1.4 spec, there are quite a few Admin commands which list "Yes" under the "Namespace Identifier Used" column. Perhaps I'm misunderstanding something, but is "ignored for admin" what we want?

imp added inline comments.Mon, Sep 21, 11:33 PM

sbin/nvmecontrol/nvmecontrol.8
769	It means that the kernel will fill in a value. This may be bogus, but it's what the kernel will do. It won't use the value we pass in. For the I/O endpoint, the kernel doesn't have this behavior. Not sure why.